A California Court Just Fined a Law Firm for Filing AI-Generated Fake Case Law. Here’s What Employers and Workers Should Take From It.

Every lawyer has cited a case without remembering the page number. Few have cited entirely fictitious quotations from cases that actually exist. It’s scary but it’s true – the use of generative AI has reached even the prestigious halls of the legal field. In a recent case, however, its use and misuse had attorneys filing a motion containing inaccurate case citations, fake quotations, and misrepresentation of controlling authority.

A recent California Court of Appeal case emphasized the importance of an attorney’s duty of competence to their client as well as to the court. In today’s world where the use of AI is almost second nature, lawyers should be cognizant to not let artificial intelligence do their thinking for them.

How a routine wage-and-hour case went sideways

The underlying dispute was ordinary enough. A class action was filed in Los Angeles County in June 2024 alleging the usual menu of wage-and-hour violations, later adding penalties under the Private Attorneys General Act (PAGA). Six months later, the same firm filed a nearly identical case in San Francisco County against the same defendants, this time with a different named plaintiff. Eight of the nine causes of action overlapped. Consequently, the defendants asked the San Francisco court to stay the second case under the doctrine of exclusive concurrent jurisdiction. The firm opposed that motion and it was in that opposition brief that this recent case has become an important reading for all lawyers using AI in their work – always read and verify the contents of the pleading before submitting it to the court.

Two fake cases, eight invented quotes

When the trial judge sat down to check the firm’s authorities, the citations didn’t hold up. Two of the cases cited as controlling law simply did not exist. Eight quotations were attributed to real cases that had never said any such thing, and in several instances the brief described case holdings as standing for the exact opposite of what they actually held.

The court concluded the brief had almost certainly been produced with a generative AI tool, and that the brief included fake quotes from genuine cases, so a quick glance at the citation list wouldn’t reveal the problem. One had to actually pull the cases and read them in order to determine the veracity of the citations.

The opinion also illustrates another important point regarding an attorney's duty of competence. Although the motion had been prepared by a contract attorney, the court emphasized that the attorney whose name appeared on the pleading remained ultimately responsible for reviewing its contents, verifying the cited authorities, and ensuring the factual representations were accurate before filing. The circumstances of the case also underscored the importance of reviewing the underlying procedural history, including whether related actions had already been filed.

The judge did not mince words at the hearing, calling it the “most egregious example of misrepresentation of the law that I’ve ever seen.” The court sanctioned the firm and three of its attorneys jointly, ordering $5,000 paid to the defendants and $1,000 paid to the court, and directed that the sanctions order be served on judges handling the related cases.

“Attorneys cannot delegate this responsibility to any form of technology”

On appeal, the firm argued it never got the 21-day “safe harbor” window that Code of Civil Procedure section 128.7 normally provides before sanctions can be imposed. The Court of Appeal didn’t reach the merits of that argument, because the firm had never timely raised it in the trial court and therefore forfeited it. The panel also noted the firm had multiple chances to withdraw the offending brief and never did.

But the part of the opinion that will get quoted for years is the court’s plain statement of the rule. As the panel put it:

The court further emphasized that neither the use of a contract attorney nor generative AI alters an attorney's accountability.

Why this matters beyond the courtroom

Although the decision arose from a particular set of facts, its significance extends beyond the parties involved. It provides guidance on the use of generative AI in legal practice and reaffirms that the duty of competence ultimately rests with the attorney, not the technology.

For employers, the case is a reminder that the technology now drafting your HR memos, offer letters, and internal investigation summaries is capable of producing confident, polished text that is simply wrong. AI tools invent citations the same way they can invent policy language, statutory deadlines, or “facts” about an employee’s file. If a document could end up in front of a judge, an agency, or an arbitrator, a human who knows the subject needs to verify it before it’s relied upon. The convenience is real; so is the exposure.

For workers, the case is a quieter kind of reassurance. Courts are paying attention to whether the lawyers on both sides of a wage-and-hour or PAGA claim are dealing squarely with the record and the law. Fabricated authority doesn’t just embarrass the firm that files it. It can delay a case, undermine otherwise legitimate claims, and cost the very employees the lawsuit was supposed to help.

The bottom line

Generative AI is a tool, not a lawyer. It is a double-edged tool. Used carefully, it can speed up research and drafting. Otherwise, it can result in avoidable errors that shift the focus away from the merits of the case and toward the accuracy of the court filings themselves. The duty of competence hasn’t changed. To exercise such competence, a lawyer must review and evaluate the output produced by generative AI before submitting it to the court. It saves everyone’s time, and honors one’s professional and ethical obligation.